This is a preliminary version of spectral clustering in C++.

The idea is to create a weighted graph from the given datapoints, where the weight between two points is given by some similarity function.
We then use properties of the eigenvectors of the graph's Laplacian to compute points that represent the data well, later to use k-means clustering on this.

A typical call is something like:

    ./SpectralClustering --clusters 3 --neighbors 4 < data.txt | ./visualization | python